cuda.tile.static_assert

cuda.tile.static_assert(condition, message=None, /)

Asserts that a condition is true at compile time.

First, condition is evaluated using the same rules as static_eval(): it can reference global and local variables, and use the full Python syntax, but must not perform any run-time operations.

The condition must evaluate to a compile-time constant boolean. If it evaluates to True, compilation continues normally, and the message expression is not evaluated.

If condition evaluates to False, then the message expression is evaluated using the static_eval() semantics. If the result of the evaluation is None, it is replaced with an empty string. Otherwise, it is converted to a string using the builtin str() function. Then, a TileStaticAssertionError is raised with the evaluated message string.

Because message is evaluated using the static_eval() semantics, it can include useful debug information about local variables, for example:

>>> x = ct.ones((4,), dtype=ct.int32)
>>> y = ct.ones((4,), dtype=ct.float32)
>>> ct.static_assert(x.dtype == y.dtype,
>>>                  f"Expected {x} and {y} to have same dtype.")
Static assertion failed: Expected <tile[int32, (4,)]> and <tile[float32, (4,)]>
 to have same dtype.

Since the message is automatically converted to a string, one can use any object in its place, for example:

>>> ct.static_assert(x.dtype == ct.float32, x)
Static assertion failed: <tile[int32, (4,)]>

Despite being declared as a function, static_assert() is treated like a keyword: it skips the translation of the surrounded expressions according to the Tile semantics. Moreover, the expressions are allowed to use the full Python syntax, unlike the rest of the Tile code, which is limited to a stricter subset of the language.